]> source.dussan.org Git - jgit.git/log
jgit.git
10 years agoGenerate javadocs as part of Maven site project reports 91/33891/1
Matthias Sohn [Thu, 25 Sep 2014 09:17:40 +0000 (11:17 +0200)]
Generate javadocs as part of Maven site project reports

see http://maven.apache.org/plugins/maven-javadoc-plugin/usage.html

Change-Id: I6dc5dddea0adbb4abf5560d5d1c833408e944082
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoCompare API changes with clirr against 3.4.1 90/33890/1
Matthias Sohn [Thu, 25 Sep 2014 09:15:57 +0000 (11:15 +0200)]
Compare API changes with clirr against 3.4.1

Change-Id: I76dcbb763984e7a7a526e29ec379d1f923b4253f
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years ago[cli] Use chaining credentials provider to enable .netrc 99/33799/3
Matthias Sohn [Tue, 23 Sep 2014 22:06:48 +0000 (00:06 +0200)]
[cli] Use chaining credentials provider to enable .netrc

If available use credentials in .netrc otherwise prompt the user
if jgit command line needs credentials for remote operations.

Change-Id: Iea1f595618edfb87a37137db08bac8ca2d500ebc
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoAdd chaining credentials provider 33/33733/4
Matthias Sohn [Wed, 17 Sep 2014 21:30:46 +0000 (23:30 +0200)]
Add chaining credentials provider

The chaining credentials provider sequentially tries to obtain
credentials from a list of credential providers and returns the
credentials from the first provider which can provide them.

Change-Id: I499f304119d7066d011dbde3556dee6facee8ab0
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years ago[Java 8] Configure doclint to accept missing descriptions 34/33734/2
Matthias Sohn [Mon, 22 Sep 2014 22:27:46 +0000 (00:27 +0200)]
[Java 8] Configure doclint to accept missing descriptions

We avoid trivial descriptions in JavaDoc. Hence configure
doclint to not fail on missing descriptions.

Change-Id: Iba3d5aec18cc7d7f43e53fa6789f0dede9996fb7
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoDo not use .netrc implicitly if no CredentialsProvider was set 10/33510/1
Matthias Sohn [Wed, 17 Sep 2014 13:16:55 +0000 (15:16 +0200)]
Do not use .netrc implicitly if no CredentialsProvider was set

Do not silently set the NetRCCredentialsProvider if no
CredentialsProvider was set explicitly since applications may want to
have full control which provider should be used.

Bug: 444338
Change-Id: Ie096983bc1caa90443a504d302bfea8f2d26ab9e
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoPrepare post 3.5.0-rc1 builds 14/33014/1
Matthias Sohn [Sun, 7 Sep 2014 23:01:31 +0000 (01:01 +0200)]
Prepare post 3.5.0-rc1 builds

Change-Id: Id9e01b867bff0ed10e887c15a12e5399b788b68e
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoJGit 3.5.0.201409071800-rc1 11/33011/1 v3.5.0.201409071800-rc1
Matthias Sohn [Sun, 7 Sep 2014 21:58:44 +0000 (23:58 +0200)]
JGit 3.5.0.201409071800-rc1

Change-Id: I59e5e2492820da111df1837fbf3649a2ae03841c
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoFix the ls-remote command when there is no local repo 93/32993/1
Robin Rosenberg [Fri, 6 Jun 2014 19:18:16 +0000 (21:18 +0200)]
Fix the ls-remote command when there is no local repo

Bug: 436695
Change-Id: I567f9a8e355c7624efa3efc9bac4f3b9015afa97
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoUpdate the README with current modules and features 52/29952/3
Robin Rosenberg [Tue, 15 Jul 2014 23:53:50 +0000 (01:53 +0200)]
Update the README with current modules and features

Change-Id: Ic30f55f30ce260dc02c428e1ebf11f01cb790a6e
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoDrop warnings about unchecked casts in a few stable select places 51/29951/3
Robin Rosenberg [Tue, 15 Jul 2014 09:37:10 +0000 (11:37 +0200)]
Drop warnings about unchecked casts in a few stable select places

Change-Id: Ie163a4940f0d13bbdefd8c4643c0944c71800544

10 years agoHide warnings about field hiding in the MyersDiff code 50/29950/3
Robin Rosenberg [Tue, 15 Jul 2014 09:36:01 +0000 (11:36 +0200)]
Hide warnings about field hiding in the MyersDiff code

Just suppress warning in this stable code rather then fixing names

Change-Id: I07580aa3cbdaa66852a5c919769bbd9020d78703

10 years agoFix warnings about missing serialVersionUID 49/29949/3
Robin Rosenberg [Tue, 15 Jul 2014 09:33:58 +0000 (11:33 +0200)]
Fix warnings about missing serialVersionUID

Use @SuppressWarnings since these are private classes

Change-Id: Ic8e4c08b70627a3f9b031e2474092e1ce8116f22

10 years agoFix warning about missing serialVersionUID 48/29948/3
Robin Rosenberg [Tue, 15 Jul 2014 09:32:45 +0000 (11:32 +0200)]
Fix warning about missing serialVersionUID

Change-Id: Ia9532d1aaf92df9073c85c44ff2ca94796deefc2

10 years agoRename local variables/parameters to remove warnings about hiding 47/29947/3
Robin Rosenberg [Tue, 15 Jul 2014 09:32:23 +0000 (11:32 +0200)]
Rename local variables/parameters to remove warnings about hiding

Change-Id: I73f38492b6a2e7fd6e77005efd0a8a8c65763e74

10 years agoRemove non-existing source directories from Java 7 test build 44/29944/3
Robin Rosenberg [Tue, 15 Jul 2014 09:27:21 +0000 (11:27 +0200)]
Remove non-existing source directories from Java 7 test build

This was a copy-artifact when the bundle was created. This chanhe
removed some warnings in Eclipse.

Change-Id: I32acdf4b60bf5528566e66b1fd82113b1343ed37

10 years agoEliminate warnings for non-nls strings that will never be translated 42/29942/3
Robin Rosenberg [Tue, 15 Jul 2014 09:22:47 +0000 (11:22 +0200)]
Eliminate warnings for non-nls strings that will never be translated

Some of these eliminations just reduces the number of warnings on
lines where messages are constructed that can/will be translated.

Change-Id: I6eddb39ccc8f2488741bb58540d9ec5f5665e2c4

10 years agoWindows: Hide the .git directory if hidedotfiles is set to non-false 83/9383/48
Robin Rosenberg [Tue, 11 Feb 2014 00:06:27 +0000 (01:06 +0100)]
Windows: Hide the .git directory if hidedotfiles is set to non-false

Other .git files are not hidden with this patch

Change-Id: Idf63ca08d08f3a77c33f5848d02074f8d6a75758
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoAdd the resources directory to the Eclipse classpath 59/29959/3
Robin Rosenberg [Wed, 16 Jul 2014 07:41:40 +0000 (09:41 +0200)]
Add the resources directory to the Eclipse classpath

This removes a warning in Eclipse.

Change-Id: I5961ebf2052aff3e5d853451f69c7d75890ac53a

10 years agoPlotCommitList: Close plot lanes at root commits 65/32665/2
Konrad Kügler [Mon, 1 Sep 2014 19:09:59 +0000 (21:09 +0200)]
PlotCommitList: Close plot lanes at root commits

Bug: 443006
Change-Id: I40b1eb044fa2d9e6c5d4b956c17e32ca6c1bd39f
Signed-off-by: Konrad Kügler <swamblumat-eclipsebugs@yahoo.de>
10 years agoPrevent NPE if no CredentialsProvider is registered 26/32626/1
Matthias Sohn [Mon, 1 Sep 2014 09:27:17 +0000 (11:27 +0200)]
Prevent NPE if no CredentialsProvider is registered

If the git server requires authentication and no CredentialsProvider is
registered TransportHttp.connect() would throw an NPE since it tries to
reset the credentials provider. Instead throw a TransportException
explaining the problem.

Change-Id: Ib274e7d9c43bba301089975423de6a05ca5169f6
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoHandle -m option for Merge command 41/32541/3
Axel Richard [Fri, 29 Aug 2014 12:37:09 +0000 (14:37 +0200)]
Handle -m option for Merge command

Set the commit message to be used for the merge commit (in case one is
created)

Bug: 442886
Change-Id: Ie5ecc13822faa366f00b3daa07f74c8441cae195
Signed-off-by: Axel Richard <axel.richard@obeo.fr>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoUploadPack: Always make PackWriter.Statistics available 76/32576/1
Shawn Pearce [Fri, 29 Aug 2014 19:56:17 +0000 (12:56 -0700)]
UploadPack: Always make PackWriter.Statistics available

If the packer fails, still obtain the stats and make them available
to the logger and the caller. Failures can frequently happen when
a client disconnects in the middle of a pack stream. Server admins
may still want to examine the timing metrics from counting and
compressing phases.

Change-Id: Iceae4f68b5473f4223d85c9edfb57837fc818eed

10 years agoPackIndexV2 should check for possible corruption 01/31501/4
Marc Strapetz [Fri, 15 Aug 2014 12:02:16 +0000 (14:02 +0200)]
PackIndexV2 should check for possible corruption

Change-Id: I1803ec6d8141f07dd4085778da6461abe81c30a9
Signed-off-by: Marc Strapetz <marc.strapetz@syntevo.com>
10 years agoAllow projects without path attribute in manifest. 18/32518/1
Yuxuan 'fishy' Wang [Fri, 29 Aug 2014 04:16:56 +0000 (21:16 -0700)]
Allow projects without path attribute in manifest.

In such case, we use the name attribute as the default value of path.

Change-Id: I53fa312d6b64c6eb2240f08af7d1d60cea99192a
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoMerge "PackIndexV2: fix possibly wrong check"
Shawn Pearce [Thu, 28 Aug 2014 23:53:36 +0000 (19:53 -0400)]
Merge "PackIndexV2: fix possibly wrong check"

10 years agoLet ObjectWalk.markUninteresting also mark the root tree as 81/31081/5
Saša Živkov [Tue, 5 Aug 2014 13:14:59 +0000 (15:14 +0200)]
Let ObjectWalk.markUninteresting also mark the root tree as
uninteresting

Using the ObjectWalk and marking a commit as uninteresting didn't mark
its root tree as uninteresting. This caused the "missing tree ..."
error in Gerrit under special circumstances. For example, if the
patch-set 2 changes only the commit message then the patch-set 1
and patch-set 2 share the same root-tree:

  ps1 -> o   o <- ps2
          \ /
           o root-tree

The transported pack will contain the ps2 commit but not the root-tree
object.

When using the BaseReceivePack.setCheckReferencedObjectsAreReachable
JGit will check the reachability of all referenced objects not provided
in the transported pack. Since the ps1 was advertised it will properly
be marked as uninteresting. However, the root-tree was reachable because
the ObjectWalk.markUninteresting missed to mark it as uninteresting.
JGit was then rejecting the pack with the "missing tree ..." exception.

Gerrit-issue: https://code.google.com/p/gerrit/issues/detail?id=1582
Change-Id: Iff2de8810f14ca304e6655fc8debeb8f3e20712b
Signed-off-by: Saša Živkov <sasa.zivkov@sap.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoHandles repo manifest file without default remote. 44/32444/1
Yuxuan 'fishy' Wang [Wed, 27 Aug 2014 23:28:43 +0000 (16:28 -0700)]
Handles repo manifest file without default remote.

Change-Id: Ic45919a0b695e8a6f4ed47cc060e0c41f1d7e7ea
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoMerge "ReceivePack: Accept shallow lines from Git >= 1.9"
Shawn Pearce [Wed, 27 Aug 2014 20:56:16 +0000 (16:56 -0400)]
Merge "ReceivePack: Accept shallow lines from Git >= 1.9"

10 years agoReceivePack: Accept shallow lines from Git >= 1.9 34/32434/3
Shawn Pearce [Wed, 27 Aug 2014 17:41:59 +0000 (10:41 -0700)]
ReceivePack: Accept shallow lines from Git >= 1.9

In Git 1.9 (5dbd767601 "support pushing from a shallow clone")
the git-core project intentionally broke the existing send-pack
protocol from shallow clients.

Shallow clients now transmit their shallow information during push,
ahead of the old-new command sequence. JGit must accept these lines
when presented.

To protect the server against clients sending partial history,
require the connectivity check when pushed to by a shallow client.

Change-Id: I46639366b0900052c376091e1688f07def44ab79

10 years agoMerge "Strip "<", ">", and "\n" from name/email in UserConfig"
Robin Rosenberg [Fri, 22 Aug 2014 13:19:49 +0000 (09:19 -0400)]
Merge "Strip "<", ">", and "\n" from name/email in UserConfig"

10 years agoMerge "Throw URISyntaxException for incorrect percent-encoding"
Robin Rosenberg [Thu, 21 Aug 2014 21:47:51 +0000 (17:47 -0400)]
Merge "Throw URISyntaxException for incorrect percent-encoding"

10 years agoStrip "<", ">", and "\n" from name/email in UserConfig 65/32065/1
Robin Stocker [Thu, 21 Aug 2014 13:10:55 +0000 (23:10 +1000)]
Strip "<", ">", and "\n" from name/email in UserConfig

This matches what C Git does, see "stripped" in `man git-commit-tree`.

It also fixes the bug of the user where an user.email like "<>" would
show up as "<<>>" in EGit.

Bug: 439844
Change-Id: I567a3c620e191ce9d37d318417e63cb5d4483419
Signed-off-by: Robin Stocker <robin@nibor.org>
10 years agoThrow URISyntaxException for incorrect percent-encoding 44/31944/3
Yi EungJun [Mon, 18 Aug 2014 09:46:19 +0000 (18:46 +0900)]
Throw URISyntaxException for incorrect percent-encoding

URIish.unescape() threw an ArrayIndexOutOfBoundsException if the given
url has incorrect percent-encoding (e.g. http://example.com/%gg). But an
URISyntaxException is much better to know the reason of the failure.

Change-Id: I3f40a26d43cd2eb4e32c11aba7dc2594bc1f98e2
Signed-off-by: Yi EungJun <eungjun.yi@navercorp.com>
10 years agoMerge "Fix minor spelling error in comment"
Robin Stocker [Tue, 19 Aug 2014 14:28:46 +0000 (10:28 -0400)]
Merge "Fix minor spelling error in comment"

10 years agoFix minor spelling error in comment 93/31893/1
Robin Rosenberg [Tue, 19 Aug 2014 14:21:02 +0000 (16:21 +0200)]
Fix minor spelling error in comment

Change-Id: I320b5cd526827844d4fe0cbc0f32551010e98090

10 years agoMerge "cleanup: Remove no-longer needed API compatibility warning filters"
Christian Halstrick [Tue, 19 Aug 2014 13:57:28 +0000 (09:57 -0400)]
Merge "cleanup: Remove no-longer needed API compatibility warning filters"

10 years agoSupport remote aliases in repo manifest. 55/31855/1
Yuxuan 'fishy' Wang [Mon, 18 Aug 2014 18:12:08 +0000 (11:12 -0700)]
Support remote aliases in repo manifest.

Change-Id: Icbe5761b9d8a4ae5305bfe45b2d042f214156fc8
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoSupport non-default remotes in project tag. 53/31853/1
Yuxuan 'fishy' Wang [Mon, 18 Aug 2014 17:57:09 +0000 (10:57 -0700)]
Support non-default remotes in project tag.

Change-Id: I3c730a1ce379d142d3ed81dda4a73f86f1f9c3eb
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoAdd tests for checking out tags with CheckoutCommand 02/31802/2
Robin Stocker [Sun, 17 Aug 2014 07:39:02 +0000 (17:39 +1000)]
Add tests for checking out tags with CheckoutCommand

Change-Id: Ifec3ac59264c91e154b832fdd907736faf8a8412
Signed-off-by: Robin Stocker <robin@nibor.org>
10 years agoAdd toString to Git with repository path 03/31803/1
Robin Stocker [Sun, 17 Aug 2014 07:48:05 +0000 (17:48 +1000)]
Add toString to Git with repository path

Useful for debugging.

Change-Id: I78b92ccbcc099370fc99a1d2fa2cf60d5f5892e6
Signed-off-by: Robin Stocker <robin@nibor.org>
10 years agoResolveMerger: Use the ObjectReader to access objects 71/31771/3
Shawn Pearce [Fri, 15 Aug 2014 17:21:32 +0000 (10:21 -0700)]
ResolveMerger: Use the ObjectReader to access objects

This is necessary to ensure objects accessed by the TreeWalk come from
the associated ObjectInserter when the merger is a RecursiveMerger
instance and a virtual common base was constructed but not flushed.

Change-Id: Iebe739d30fd868ebc4f61dbfb714673146a2c3ec

10 years agoPackIndexV2: fix possibly wrong check 52/31752/1
Marc Strapetz [Fri, 15 Aug 2014 12:00:40 +0000 (14:00 +0200)]
PackIndexV2: fix possibly wrong check

According to http://stackoverflow.com/a/8381338, the maximum array
size is not Integer.MAX_VALUE, but Integer.MAX_VALUE - 8

Change-Id: I6ddc7470368acd20abf0885c53c89a982bb0f176
Signed-off-by: Marc Strapetz <marc.strapetz@syntevo.com>
10 years agoCleanup use of java.util.Inflater, fixing rare infinite loops 04/31604/4
Shawn Pearce [Wed, 13 Aug 2014 23:54:22 +0000 (16:54 -0700)]
Cleanup use of java.util.Inflater, fixing rare infinite loops

The native implementation of inflate() can set finished to return
true at the same time as it copies the last bytes into the buffer.

Check for finished on each iteration, terminating as soon as libz
knows the stream was completely inflated.

If not finished, it is likely input is required before the next
native call could do any useful work. Most invocations are passing
in a buffer large enough to store the entire result. A partial return
from inflate() will need more input before it can continue. Checking
right away that needsInput() is true saves a native call to determine
no bytes can be inflated without more input.

This should fix a rare infinite loop condition inside of inflation
when an object ends exactly at the end of a block boundary, and
the next block contains only the 20 byte trailing SHA-1.

When the stream is finished each new attempt to inflate() returns
n == 0, as no additional bytes were output. The needsInput() test
tries to add the length of the footer block to itself, but then loops
back around an reloads the same block as the block is smaller than
a full block size. A zero length input is set to the inflater,
which triggers needsInput() condition again.

Change-Id: I95d02bfeab4bf995a254d49166b4ae62d1f21346

10 years agoAdd missing @since 3.5 tag to RepoCommand.setIncludedFileReader 98/31598/1
Shawn Pearce [Wed, 13 Aug 2014 22:28:24 +0000 (15:28 -0700)]
Add missing @since 3.5 tag to RepoCommand.setIncludedFileReader

Change-Id: Ib5d24f8dafacc0d12c36927c2bbc3ebef2fff7b2

10 years agoAdd support to <include> tag in repo manifest xml. 50/31150/4
Yuxuan 'fishy' Wang [Thu, 7 Aug 2014 05:01:08 +0000 (22:01 -0700)]
Add support to <include> tag in repo manifest xml.

Change-Id: I32d468f92e24701ea680435bf3417e3850857303
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoMerge "Tweak the handling of remote url in jgit repo."
Shawn Pearce [Wed, 13 Aug 2014 20:02:57 +0000 (16:02 -0400)]
Merge "Tweak the handling of remote url in jgit repo."

10 years agoRevert "Add a method to DfsOutputStream to read as an InputStream" 52/31552/1
Shawn Pearce [Wed, 13 Aug 2014 17:42:43 +0000 (10:42 -0700)]
Revert "Add a method to DfsOutputStream to read as an InputStream"

This reverts commit b646578d89513c5db30151f0b865c7335c88d682.

openInputStream() is never used in JGit, nor is it used by any
known working DFS implementation. The method was added as a
utility for reading back from a DfsInserter, but the final
implementation of that feature does not requrire this method.

Change-Id: I075ad95e40af49c92b554480f8993ef5658f7684

10 years agoAdd a Merger.merge method that makes flushing optional 04/12504/8
Dave Borowitz [Fri, 3 May 2013 18:45:36 +0000 (11:45 -0700)]
Add a Merger.merge method that makes flushing optional

This allows callers performing multiple separate merges to reuse a
single ObjectInserter without flushing the inserter on each iteration
(which can be slow in the DFS case).

Change-Id: Icaff7d2bc2c20c873ce5a7d9af5002da84ae1c2b

10 years agoTeach mergers to use ObjectInserter.newReader() 03/12503/10
Dave Borowitz [Fri, 3 May 2013 17:54:38 +0000 (10:54 -0700)]
Teach mergers to use ObjectInserter.newReader()

This allows the RecursiveMerger to iteratively create new merge bases
without necessarily flushing packs to storage in the DFS case;
flushing only need happen at the end of the whole merge process.

Since Merger's walk now depends on its inserter, we need to construct
an inserter at Merger construction time.  This should not be a
significant increase in overhead since unused inserters don't use any
resources (beyond a reference to the Repository).

We also must release and recreate the walk whenever setObjectInserter
is called, which can break usages where setObjectInserter is called in
the middle of stateful operations on the walk.  No usages of this
method within JGit currently do this; the inserter is only ever set
before any stateful walk operations happen.

Change-Id: I9682a6aa4a2c37dccef8e163f132ddb791d79103

10 years agoTweak the handling of remote url in jgit repo. 62/31562/1
Yuxuan 'fishy' Wang [Wed, 13 Aug 2014 17:36:56 +0000 (10:36 -0700)]
Tweak the handling of remote url in jgit repo.

This kind of reverted 31148. URI.resolve actually can handle the absolute URL
well, the problem is only the missing "/".

Change-Id: Iee5866c005cbc1430dc20ee7db321b8b51afed30
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoAdd a method to ObjectInserter to read back inserted objects 22/12422/11
Dave Borowitz [Wed, 1 May 2013 17:20:31 +0000 (10:20 -0700)]
Add a method to ObjectInserter to read back inserted objects

In the DFS implementation, flushing an inserter writes a new pack to
the storage system and is potentially very slow, but was the only way
to ensure previously-inserted objects were available.  For some tasks,
like performing a series of three-way merges, the total size of all
inserted objects may be small enough to avoid flushing the in-memory
buffered data.

DfsOutputStream already provides a read method to read back from the
not-yet-flushed data, so use this to provide an ObjectReader in the
DFS case.

In the file-backed case, objects are written out loosely on the fly,
so the implementation can just return the existing WindowCursor.

Change-Id: I454fdfb88f4d215e31b7da2b2a069853b197b3dd

10 years agoDfsInserter: buffer up to streamFileThreshold from InputStream 82/31482/3
Shawn Pearce [Tue, 12 Aug 2014 21:57:28 +0000 (14:57 -0700)]
DfsInserter: buffer up to streamFileThreshold from InputStream

Since 2badedcbe0f87c0a in-core merges can write up to 10 MiB
into a TemporaryBuffer.Heap strategy, where the data is stored
as a chain of byte[] blocks.

Support the inserter reading up to the streamFileThreshold (default 50
MiB) from the supplied input stream and hash the content to determine
if the merged result blob is already present in the repository.  This
allows the inserter to avoid creating duplicate objects in more cases,
reducing repository pack file churn.

Change-Id: I38967e2a0cff14c0a856cdb46a2c8fedbeb21ed5

10 years agoMerger: clarify getBaseCommit javadoc 02/12502/3
Dave Borowitz [Fri, 3 May 2013 17:21:50 +0000 (10:21 -0700)]
Merger: clarify getBaseCommit javadoc

Change-Id: Ic52475248a545d51d3d8ac31a24f0344a1241e45

10 years agoReuse Merger's ObjectReader from ResolveMerger 01/12501/7
Dave Borowitz [Wed, 1 May 2013 22:47:23 +0000 (15:47 -0700)]
Reuse Merger's ObjectReader from ResolveMerger

The base Merger class already has a single ObjectReader instance that
it handles releasing as necessary, so creating new readers is not
necessary.

Change-Id: I990ec43af7df448c7825fc1b10e62eadaa3e0c2a

10 years agoMerge "Fix CheckoutCommand not setting up tracking"
Robin Rosenberg [Mon, 11 Aug 2014 20:23:38 +0000 (16:23 -0400)]
Merge "Fix CheckoutCommand not setting up tracking"

10 years agoMerge "Process most in-core merges without local temp files"
Shawn Pearce [Mon, 11 Aug 2014 16:42:10 +0000 (12:42 -0400)]
Merge "Process most in-core merges without local temp files"

10 years agoMerge "Update tycho and signing plugins to latest versions"
Robin Rosenberg [Mon, 11 Aug 2014 15:26:21 +0000 (11:26 -0400)]
Merge "Update tycho and signing plugins to latest versions"

10 years agoMerge "Update maven plugins to latest versions"
Robin Rosenberg [Mon, 11 Aug 2014 15:25:57 +0000 (11:25 -0400)]
Merge "Update maven plugins to latest versions"

10 years agoUpdate tycho and signing plugins to latest versions 30/31330/3
Matthias Sohn [Fri, 8 Aug 2014 22:46:43 +0000 (00:46 +0200)]
Update tycho and signing plugins to latest versions

Change-Id: I3e8b3835b8ebbb8f0a2a9be42e6c16899cb3b993
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoUpdate maven plugins to latest versions 29/31329/3
Matthias Sohn [Fri, 8 Aug 2014 22:44:21 +0000 (00:44 +0200)]
Update maven plugins to latest versions

Also, remove unused findbugs exclude filter in java7 bundle since latest
findbugs plugins raises an error on this.

Change-Id: I791fc054596e7d9aa9f3cc8126eb0162539c57bf
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoMerge "Remove redundant semicolon"
Robin Rosenberg [Sun, 10 Aug 2014 08:22:08 +0000 (04:22 -0400)]
Merge "Remove redundant semicolon"

10 years agoMerge "We should not warn for unqualifiedFieldAcces"
Robin Rosenberg [Sun, 10 Aug 2014 08:21:55 +0000 (04:21 -0400)]
Merge "We should not warn for unqualifiedFieldAcces"

10 years agoReplace deprecated method call in Java7 FTI test 46/29946/3
Robin Rosenberg [Tue, 15 Jul 2014 09:29:57 +0000 (11:29 +0200)]
Replace deprecated method call in Java7 FTI test

Change-Id: I3bbb6da951765d79d77aa955edeac7b5e4da5d8a

10 years agoWe should not warn for unqualifiedFieldAcces 43/29943/3
Robin Rosenberg [Tue, 15 Jul 2014 09:25:17 +0000 (11:25 +0200)]
We should not warn for unqualifiedFieldAcces

It does not make sense and I cannot explain why it slipped through.

Change-Id: I005a54cde545ea99553c0a2d6dd8d64e6e0a7611

10 years agoRemove redundant semicolon 45/29945/3
Robin Rosenberg [Tue, 15 Jul 2014 09:28:50 +0000 (11:28 +0200)]
Remove redundant semicolon

Change-Id: I15370d7807c82ee85ed7fdb05061a4baf0a77a68

10 years agoProcess most in-core merges without local temp files 36/31336/1
Shawn Pearce [Sat, 9 Aug 2014 00:53:49 +0000 (17:53 -0700)]
Process most in-core merges without local temp files

Instead of always writing to disk use TemporaryBuffer.LocalFile to
store up to 10 MiB of merge result in RAM. Most source code will
fit into this limit, avoiding local disk IO for simple merges.

Larger files will automatically spool to a temporary file that
can be cleaned up in the finally, reducing the risk of leaving
them on disk and consuming space in /tmp.

Change-Id: Ieccbd9b354d4dd3d2bc1304857325ae7a9f34ec6

10 years agoMerge "ResolveMerger: push result file creation into updateIndex()"
Matthias Sohn [Fri, 8 Aug 2014 22:34:47 +0000 (18:34 -0400)]
Merge "ResolveMerger: push result file creation into updateIndex()"

10 years agoResolveMerger: push result file creation into updateIndex() 26/31326/1
Shawn Pearce [Fri, 8 Aug 2014 22:04:57 +0000 (15:04 -0700)]
ResolveMerger: push result file creation into updateIndex()

The only caller of writeMergedFile is updateIndex, and the only
user of this path object is the code within the method. This is
a no-op change that opens the door to refactoring the way temp
files are handled for inCore merges.

Change-Id: I863a303194689a806b667e55eb958e1decf046c1

10 years agoUpdate Maven JAR plugins 12/31312/1
Shawn Pearce [Fri, 8 Aug 2014 18:26:16 +0000 (11:26 -0700)]
Update Maven JAR plugins

Seems to fix the build stalls observed under some builds of the
Java 7 JRE.

Change-Id: I94d5e927b5b2ec3971e9a70c18966368a61ec14c

10 years agoRecursiveMerger: do not insert virtual commits into repository 11/31311/2
Shawn Pearce [Fri, 8 Aug 2014 18:19:45 +0000 (11:19 -0700)]
RecursiveMerger: do not insert virtual commits into repository

When merging common ancestors to create a single virtual common
ancestor the commit does not need to be inserted into the Git
repository. Instead just mock it out in memory as part of the
merger's RevWalk pool.

Make the author and committer stable and predictable for any
given pair of merge bases. It is not necessary for the caller's
name or email to be used as the commit will not be written out.

Change-Id: I88d5ee4de121950e1b032a5c10486c9d2c42656c

10 years agoFix non-relative remote defined in manifest xml. 48/31148/3
Yuxuan 'fishy' Wang [Thu, 7 Aug 2014 01:17:29 +0000 (18:17 -0700)]
Fix non-relative remote defined in manifest xml.

Currently if the remote defined in repo manifest xml is non-relative (e.g.
"https://chromium.googlesource.com"), our code will break. This change fixed
that.

It also makes that remotes are ending with "/".

Change-Id: Icef46360b32227a9db1d9bb9e6d929c72aeaa8df
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoFix CheckoutCommand not setting up tracking 42/31042/1
Robin Stocker [Tue, 5 Aug 2014 13:38:07 +0000 (23:38 +1000)]
Fix CheckoutCommand not setting up tracking

Instead of passing on the start point as is to CreateBranchCommand, the
resolved ObjectId was used. Given this, CreateBranchCommand did not set
up tracking.

This also fixes CreateBranchCommand with setStartPoint(null) to use HEAD
(instead of NPEing), as documented in the Javadoc.

Bug: 441153
Change-Id: I5ed82b4a4b4a32a81a7fa2854636b921bcb3d471
Signed-off-by: Robin Stocker <robin@nibor.org>
10 years agoRun unit tests in parallel 36/30936/1
Matthias Sohn [Sun, 3 Aug 2014 23:04:26 +0000 (01:04 +0200)]
Run unit tests in parallel

Configure Maven surefire plugin to run tests concurrently using
multiple processes. By default use one process per core available
on the machine running the tests.

Running this on my MacBook Pro i7 with 8 cores 16 GB RAM SSD
I get the following results:

forkCount   "mvn clean install" on org.eclipse.jgit.test
--------------------------------------------------------
1           02:36 min
2           01:26 min
4           00:56 min
6           00:49 min
8           00:49 min
10          00:51 min

When on-access scan of McAfee virus scanner is enabled the optimum
is at half the number of cores (4 of 8) on my MacBook.

Set system property "test-fork-count" to override the default.
Either set it to an absolute number e.g. "4" or in units of
cores available on the system running the build. E.g. "0.5C" to
use 4 cores on a 8 core system or "1C" to use all 8 cores on
a 8 core system.
E.g.
$ mvn clean install -Dtest-fork-count=0.5C
to use half the number of cores processes to run tests.

Change-Id: Ifdbd1ae8153f2033922eae6ae8942c36db1f9806
Signed-off-by: Sohn <matthias.sohn@sap.com>
10 years agoMerge "Add isRebase to API of BranchConfig"
Robin Rosenberg [Sun, 3 Aug 2014 11:33:28 +0000 (07:33 -0400)]
Merge "Add isRebase to API of BranchConfig"

10 years agoMerge "Use bitcheck to check for presence of OPT_FULL option"
Robin Rosenberg [Sun, 3 Aug 2014 08:51:12 +0000 (04:51 -0400)]
Merge "Use bitcheck to check for presence of OPT_FULL option"

10 years agoAdd isRebase to API of BranchConfig 24/30924/1
Robin Stocker [Sun, 3 Aug 2014 08:32:43 +0000 (18:32 +1000)]
Add isRebase to API of BranchConfig

Change-Id: I9819f49410e30d32c2157db0556a0dd6a0bcc5a4
Signed-off-by: Robin Stocker <robin@nibor.org>
10 years agoNetRC parser and credentials provider. 94/22194/8
Alexey Kuznetsov [Tue, 18 Feb 2014 22:23:09 +0000 (02:23 +0400)]
NetRC parser and credentials provider.

Adds default git command line behaviour of reading credentials
for https connections from .netrc file.

Bug: 428229
Change-Id: I88699ca5da6a20bdeaed24b7e3899cc1022b8e7c
Signed-off-by: Alexey Kuznetsov <axet@me.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agocleanup: Remove no-longer needed API compatibility warning filters 41/29941/2
Robin Rosenberg [Tue, 15 Jul 2014 09:16:54 +0000 (11:16 +0200)]
cleanup: Remove no-longer needed API compatibility warning filters

Change-Id: If534bd0aedaee6160679e965ad08ac30d5dac1c5

10 years agoUse bitcheck to check for presence of OPT_FULL option 94/17894/3
Kevin Sawicki [Wed, 30 Oct 2013 15:55:14 +0000 (08:55 -0700)]
Use bitcheck to check for presence of OPT_FULL option

Previously an equality check was performed so an exception would
be thrown if any other options were set.

Change-Id: I36b60e2c0a8aef9fcfe663055dba520192996872

10 years agoEnhance BranchConfig API to expose branch.* options 15/30915/2
Robin Stocker [Sat, 2 Aug 2014 09:20:51 +0000 (19:20 +1000)]
Enhance BranchConfig API to expose branch.* options

With these, more code can use BranchConfig instead of directly accessing
the raw configuration values.

Change-Id: I4b52f97ff0e3fc8f097512806f043c615a3d2594
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoAdd IndexDiff tests for merge conflict state BOTH_ADDED 85/30585/5
Laurent Goubet [Mon, 28 Jul 2014 12:27:05 +0000 (14:27 +0200)]
Add IndexDiff tests for merge conflict state BOTH_ADDED

JGit handled this case improperly which these tests demonstrate. Fixed
by I25915880f304090fe90584c79bddf021231227a2.

Bug: 440537
Change-Id: Ia29c1d6cf8c0ce724cc3ff5ed9e0b396949b44bf
Signed-off-by: Laurent Goubet <laurent.goubet@obeo.fr>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoIndexDiffFilter should never filter entries with stage!=0 58/30858/2
Christian Halstrick [Fri, 1 Aug 2014 09:28:01 +0000 (11:28 +0200)]
IndexDiffFilter should never filter entries with stage!=0

If the IndexDiffFilter is asked whether it should include or filter out
a certain path and for that path there is a dircache entry with a stage
different from 0, then the filter should never filter out this entry.
IndexDiffFilter is an optimized version of AnyDiffFilter and there is no
case where the index contains non-0 stages but we still don't see any
diff for that path.

Change-Id: I25915880f304090fe90584c79bddf021231227a2
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoMerge "HistogramDiff: Convert stack recursion to heap managed queue"
Robin Rosenberg [Sat, 2 Aug 2014 09:35:18 +0000 (05:35 -0400)]
Merge "HistogramDiff: Convert stack recursion to heap managed queue"

10 years agoFix Javadoc errors with Java 8 49/30549/2
Robin Stocker [Sun, 27 Jul 2014 05:13:36 +0000 (15:13 +1000)]
Fix Javadoc errors with Java 8

Bug: 440414
Change-Id: Icce87bd23a3698513b76246677f1d45ddbcaab47
Signed-off-by: Robin Stocker <robin@nibor.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoFix API errors raised on ResolveMerger affecting API providers only 67/30767/2
Christian Halstrick [Wed, 30 Jul 2014 15:01:56 +0000 (17:01 +0200)]
Fix API errors raised on ResolveMerger affecting API providers only

In change If45bc3d078b3d3de87b758e71d7379059d709603 a new parameter was
added to 3 protected methods of ResolveMerger. This breaks the code of
developers which have subclassed ResolveMerger. The API baseline check
in Eclipse reports this as API breakage.

Since this will break only providers but not consumers of the API this
should be allowed also in minor versions. According to OSGi semantic
versioning
http://www.osgi.org/wiki/uploads/Links/SemanticVersioning.pdf
breaking providers in a minor version update is ok.

Therefore silence these errors using API filter rules.

Bug: 440757
Change-Id: Icabbd0e1de7e877c66a5c4a2c8391473f992a1aa
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoImplement RepoCommand.Project.hashCode() since we override equals() 66/30766/2
Matthias Sohn [Wed, 30 Jul 2014 21:10:13 +0000 (23:10 +0200)]
Implement RepoCommand.Project.hashCode() since we override equals()

Bug: 440757
Change-Id: I16614c75b62b2fbef05419dd8667e3c50889abe4
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoRemove @Override annotation illegal for interfaces on Java 5 65/30765/2
Matthias Sohn [Wed, 30 Jul 2014 21:07:13 +0000 (23:07 +0200)]
Remove @Override annotation illegal for interfaces on Java 5

Bug: 440757
Change-Id: I11897e8e680722313f28d9d241e23e92eddeb966
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoMark package org.eclipse.jgit.gitrepo.internal as x-internal 64/30764/2
Matthias Sohn [Wed, 30 Jul 2014 21:05:33 +0000 (23:05 +0200)]
Mark package org.eclipse.jgit.gitrepo.internal as x-internal

It's an internal package which isn't part of the API. Mark it x-internal
to silence @since tag warnings which are only raised for new API.

Bug: 440757
Change-Id: Id05deaca43f135cd1bfe83cf1f29787cbbdbecac
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoSmall performance optimization for ignore rules/fnmatcher 27/30727/2
Andrey Loskutov [Wed, 30 Jul 2014 08:42:30 +0000 (10:42 +0200)]
Small performance optimization for ignore rules/fnmatcher

 - don't check empty segments generated by String.split()
 - don't continue to add segments if the matcher fails to match input
 - don't add empty heads
 - don't iterate over empty heads.

Bug: 440732
Change-Id: I7d04dccfe24d91275d17ba246662337d6dba66df
Signed-off-by: Andrey Loskutov <loskutov@gmx.de>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoAdded toString() methods for better debugging of ignore rules 26/30726/2
Andrey Loskutov [Wed, 30 Jul 2014 08:38:53 +0000 (10:38 +0200)]
Added toString() methods for better debugging of ignore rules

Change-Id: Ie31687faa2df47ecaacace2504c3b4e93f1ea809
Signed-off-by: Andrey Loskutov <loskutov@gmx.de>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoFix RevertCommand to correctly revert multiple commits at once. 26/30026/3
Maik Schreiber [Thu, 19 Jun 2014 10:54:14 +0000 (12:54 +0200)]
Fix RevertCommand to correctly revert multiple commits at once.

The fix is to move the new head commit to the newly-created revert
commit, so that additional revert commits will use the correct head.

Change-Id: I5de3a9a2a4c276e60af732e9c507cbbdfd1a4652
Signed-off-by: Maik Schreiber <blizzy@blizzy.de>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
10 years agoMerge "Remove overlapping submodules from repo manifest."
Shawn Pearce [Tue, 29 Jul 2014 23:28:11 +0000 (19:28 -0400)]
Merge "Remove overlapping submodules from repo manifest."

10 years agoRemove overlapping submodules from repo manifest. 99/30699/2
Yuxuan 'fishy' Wang [Tue, 29 Jul 2014 22:43:04 +0000 (15:43 -0700)]
Remove overlapping submodules from repo manifest.

Apparently repo allows projects overlapping, e.g. one project's path is "foo"
and another project's path is "foo/bar". This is not supported in git submodule.
At JGit repo side we'll skip all the submodules that are in subdirectories of
other submodules, and on repo side we'll make them submodules to resolve this
problem.

Change-Id: I6820c4ef400c530a36150b1228706adfcc43ef64
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoMerge "Added the missing export in MANIFEST.MF."
Shawn Pearce [Tue, 29 Jul 2014 23:01:13 +0000 (19:01 -0400)]
Merge "Added the missing export in MANIFEST.MF."

10 years agoMerge "RecursiveMerger should not fail on content-merge conflicts of parents"
Christian Halstrick [Tue, 29 Jul 2014 07:46:52 +0000 (03:46 -0400)]
Merge "RecursiveMerger should not fail on content-merge conflicts of parents"

10 years agoStrip trailing /s in repo base URI. 16/28216/7
Yuxuan 'fishy' Wang [Mon, 9 Jun 2014 18:27:52 +0000 (11:27 -0700)]
Strip trailing /s in repo base URI.

Change-Id: I71040db2cca69638cfad89acdc9179ee80592b27
Signed-off-by: Yuxuan 'fishy' Wang <fishywang@google.com>
10 years agoRecursiveMerger should not fail on content-merge conflicts of parents 62/30562/2
Christian Halstrick [Fri, 25 Jul 2014 06:41:38 +0000 (08:41 +0200)]
RecursiveMerger should not fail on content-merge conflicts of parents

Previously when RecursiveMerger was trying to create a single virtual
common base for the merge it was failing when this lead to content-merge
conflicts. This is different from what native git is doing. When native
git's recursive merge algorithm creates a new common base it will merge
the multiple parents and simply take the merge result (potentially
including conflict markers) as common base. See my discussion with Shawn
here: http://www.spinics.net/lists/git/msg234959.html :

  > - How should workingtree, index (stage1,2,3) look like if during
that
  > merge of common ancestors a conflict occurs? Will I see in stage2
and
  > stage3 really see content of X1 and X2?

  Its done entirely in memory and never touches the working tree or
  index. When a conflict exists in the X1-X2 merge the conflict is
  preserved into the new virtual base.

There is still the possibility that the merge of parents lead to
conflicts. File/Folder conclicts, conflicts on filemodes. This commit
only fixes the situation for conflicts when merging content.

Bug: 438203
Change-Id: If45bc3d078b3d3de87b758e71d7379059d709603

10 years agoFixed message for exception thrown during recursive merge 61/30561/2
Christian Halstrick [Thu, 26 Jun 2014 11:42:21 +0000 (13:42 +0200)]
Fixed message for exception thrown during recursive merge

During recursive merge jgit potentially has to merge multiple
common ancestors. If this fails because there are conflicts then
the exception thrown for that should have a message which states
this clearly. Previously a wrong message was given ("More than 200
merge bases ...")

Change-Id: Ia3c058d5575decdefd50390ed83b63668d31c1d1

10 years agoHistogramDiff: Convert stack recursion to heap managed queue 26/30526/4
Shawn Pearce [Fri, 25 Jul 2014 15:17:27 +0000 (08:17 -0700)]
HistogramDiff: Convert stack recursion to heap managed queue

Each time the longest common substring is found the diff algorithm
recurses to reprocess the regions before and after the common string.

Large files with many edits can trigger StackOverflowError as the
algorithm attempts to process a deeply split tree of regions. This
is especially prone to happen in servers where the Java stack size
may have been limited to 1M or even 256K.

To keep edits produced in order a queue is used to process edits
in a depth-first strategy.

Change-Id: Iae7260c6934efdffac7c7bee4d3633a8208924f7